Old English Verb-Second-ish in a Typology of Verb-Second

Benjamin Bruening, University of Delaware, bruening@udel.edu

draft November 9, 2016, comments welcome

Abstract

Old English main clauses were often verb-second (V2), but sometimes they departed from strict V2, in ways thatwere very unlike all the modern V2 languages. Old English word order is therefore puzzling and has resisteda good analysis, and it is also problematic for theories of the typology of V2 across languages. I argue that theCP Recursion model proposed in Bruening (2016) for the peculiarities of Present-Day English subject-auxiliaryinversion and cross-linguistic variation provides a new perspective on Old English V2-ish. From this perspective,Old English is just the combination of Present-Day English and modern V2 languages like German. The onlyhistorical change is the loss of the German option. This perspective therefore situates Old English nicely withina typology of V2 and the left periphery generally, and requires minimal historical change to get from Old Englishto Present-Day English.

1 Introduction

The manifestation of verb-second (V2) word order in Old English was very different from any other known V2language. The peculiarities of Old English word order have so far resisted a satisfactory account, and Old Englishcannot be located comfortably within existing typologies of V2 across languages. In this paper I show that OldEnglish can be fit nicely within a simple typology of V2, namely, the system of CP recursion plus V-C Alignmentproposed in Bruening (2016). Within this system, Old English can be seen to have two options: the Present-DayEnglish setting, or the setting that is common among modern Germanic languages (for instance German). We thussee that there is nothing unusual at all about Old English: it is located squarely between Present-Day English andthe more common Germanic pattern, exactly where we would expect it to be. There is also much more historicalcontinuity than previously thought: as I will show, many of the facts of Old English word order are exactly those ofPresent-Day English. The major change that we find is the loss of the common Germanic option, and the restrictionto the Present-Day English setting. This plus other changes that we know independently took place (e.g., loss ofmain verb movement, restriction of V2 to certain contexts, etc.) lead to the Present-Day English grammar. Thisperspective therefore leads to a maximally simple account of language change, and a simple and restrictive typologyof V2 that includes within it the apparently divergent Old English.

I begin by outlining the main facts of Old English word order (section 2). Section 3 discusses some previousanalyses that have been proposed and how they are less than satisfactory. Section 4 then outlines the CP recursiontheory proposed in Bruening (2016), and derives a typology of V2 from this theory. Section 5 then looks at OldEnglish from the perspective of this typology, and shows how the facts are the combination of the German andPresent-Day English settings. This section also outlines a proposed route of change, where minimal changes leadfrom Old English to Present-Day English.

2 Old English Word Order

In many cases, word order in Old English seems to be very similar to that of the modern OV Germanic languages,like German. Main clauses are often V2, as the following examples with fronted elements other than the subject

1

show:1

(1) a. Ælceach

rihtright

sculonmust

gehadodemonastic

menmen

lufianlove

(WHom 10a.10)

‘Monastic men must love each right. . . ’b. þæt

thathusbuilding

hæfdonhad

RomaneRomans

towith

ðæmthe

anumone

tacnefeature

geworht. . . (Orconstructed

59.3)

‘The Romans had constructed that building with the one feature. . . ’c. þær

therewearþwas

sethe

cyningking

BagsecgBagsecg

ofslægenslain

(ChronA 871)

‘The king Bagsecg was slain there.’d. maran

morecyððeaffinity

habbaðhave

englasangels

toto

GodeGod

þonnethan

men.men

(AECHom I 1,14)

‘angels have more affinity to God than men’e. Ac

andeallall

ðisthis

aredaðarranges

sethe

reccereruler

swiðevery

ryhterightly

(CP 168,3)

‘and the ruler arranges all this very rightly’f. On

intwamtwo

þingumthings

hæfdehad

GodGod

þæsthe

mannesman’s

sawlesoul

gegodod.endowed

(AECHom I 1,161)

‘God had endowed man’s soul with two things.’

Embedded clause are often verb-final, and often have the order main-auxiliary if there is more than one verb:

(2) a. . . . gifif

hiethey

himhim

þæsthe

riceskingdom

uþon,granted

(ChronA 755)

‘. . . if they would grant him the kingdom,’b. . . . and

andaxodeasked

hwiwhy

hehe

hishis

bebodcommandment

tobræcebroke

(AECHom I 1,42)

‘. . . and asked why he had broken his commandment’c. Soðlice

indeedæfterafter

þam þewhen

apolloniusApollonius

afarengone

wæs,was

(ApT 6.1)

‘Indeed when Apollonius had gone, . . . ’d. . . . þæt

thathiethey

þætit

tofrom

hishis

hondahand

healdanhold

sceoldon.should

(ChronA 887)

‘. . . that they ought to hold it from him as overlord.’e. þæt

thathithey

micclumgreatly

blissianrejoice

mihton. . . (AECHommight

I 26,279)

‘that they might greatly rejoice’

But note that the order auxiliary-main is also common, and it is also common for constituents to follow the verb.See especially Pintzuk (1999) for discussion.

So far this is just like modern German. Unlike modern German, however, Old English also allowed V2 inclauses embedded beneath a complementizer. By far the most common constituent before the finite verb in thesecases is the subject, but such embedded clauses appear to have fronting of a verb from an underlying verb-finalstructure because of the order of object and main verb in (3a–b) and the position of the stranded particle in (3c).There is also occasional fronting of a non-subject, as in (3d–f):

1Old English texts are cited using the short titles employed by the Dictionary of Old English (Healey and Venezky 1980). Texts are citedby chapter (and page number) and line number or just line number, except that Or, CP, and GD are cited by page number rather than chapternumber.

2

(3) a. andand

cwæðsaid

þthat

wewe

sceoldonshould

deaðedeath

sweltanperish

gif. . . (AECHomif

I 1,132)

‘and said that we should perish by death if. . . ’b. and

andseothe

modormother

behetpromised

himhim

þætthat

heoshe

woldewould

hinehim

læranteach

(AELS 25.173)

‘and the mother promised him that she would teach him.’c. forþon

becausenenot

cymðcomes

nahtin-no-way

ungelicdifferent

trymnesconfirmation

uppup

(GD(C) 8,1)

‘. . . because a different confirmation in no way comes up. . . ’d. Ða

thencwæðsaid

sethe

halgaholy

bisceopbishop

þætthat

onin

þamthat

beametree

nærenot-were

nanno

synderlicspecial

halignyssholiness

(AELS 31,396)

‘Then the holy bishop said that there was no special holiness in that tree’e. Gregorius

Gregorysethe

trahtnereinterpreter

cwæðsaid

þætthat

for ðitherefore

woldewanted

drihtenGod

getrahtnianinterpret

þurhthrough

hine sylfnehimself

þætthe

bigspelparable

ðethat

(AECHom II,6,33)

‘Gregory the interpreter said that therefore God wanted to interpret himself the parable that. . . ’f. ðam

theaðeoath

þætwhich

hinehim

motonmight

hishis

mægasrelatives

unsyngian.exculpate

(LawIne 65,8)

‘. . . the oath by which his relatives might absolve him.’

See especially Pintzuk (1993, 1999) on embedded V2 in Old English.While embedded V2 is not allowed in German, it is allowed in modern Icelandic and Yiddish (Platzack 1986,

Rögnvaldsson and Thráinsson 1990, Diesing 1990, Santorini 1992), and to a lesser extent in Frisian and Danish(Vikner 1995). So far, then, Old English is not unusual from the perspective of modern V2 languages.2

However, unlike all known V2 languages, Old English also sometimes exhibits V3 or even V4 order in mainclauses (and occasionally in embedded clauses). The primary (but not exclusive) context for this involves pronom-inal arguments. Pronouns tend to appear between the first constituent and the finite verb, making the finite verbthird or fourth:

(4) a. Ælceach

yfelevil

hehe

mægcan

don,do

(WHom, 4,62)

‘He can do each evil,’b. Scortlice

brieflyicI

hæbbehave

nunow

gesædspoken

ymbabout

þathe

þriethree

dælas. . . (Orparts

9,18)

‘I have now spoken briefly about the three parts.’c. þin

thineagenown

geleafafaith

þethee

hæfþhas

gehælednehealed

(HomS 2,27)

‘Thine own faith hath made thee whole.’d. &

andseofonseven

ærendracanmessengers

hehe

himhim

hæfdehad

toto

asend.send

(ChronA 905)

‘He had to send seven messengers to him.’2According to Fischer et al. (2000, 109), 95% of main clauses in Old English had verb fronting (V2 or V3/V4), while only 35% of

embedded clauses were V2 (with V3 very rare in embedded clauses). Note also that by far the most frequent initial constituent in embeddedV2 is the subject of the embedded clause (non-subjects front 6.2% of the time, according to Pintzuk 1993, 21). This has led some authorsto consider embedded V2 a distinct phenomenon from main clause V2 (e.g, Fischer et al. 2000, 109). However, I take this to indicate onlythat fronting of non-subjects is much less frequent in embedded clauses, and fronting of subjects is much more common. Otherwise, theanalysis remains the same; see section 5.

3

e. Forðontherefore

wewe

sceolanshall

midwith

ealleall

modmind

&and

mægenepower

toto

GodeGod

gecyrranturn

(HomU 8,27)

‘Wherefore we must with all mind and might turn to God’

Pronouns can also appear before a fronted constituent, again making the verb third or fourth:

(5) a. &and

icI

gehwameveryone

willewill

þærtothereto

tæcan. . . (Ordirect

57.15)

‘. . . and I will direct everyone there. . . ’b. he

hehinehim

eftagain

ongonbegan

wætereswater

weorpan,sprinkle

(Beo 2790–2791)

‘He began to sprinkle him again with water. . . ’

These verb-third and verb-fourth orders are what distinguish Old English from all other V2 languages, and are theprimary topic of the present paper.3

It should also be noted that in three particular contexts, V2 is dominant, and V3 and V4 very rare. These are,first, wh-questions:

(6) a. Hwætwhat

sægestsay

þuyou

yrþlingc?ploughman

Huhow

begæstdo

þuyou

weorcwork

þin?your

(AEColl 22)

‘What do you say, ploughman? How do you go about your work?’b. hwi

whysceoleshould

wewe

oþresanother

mannesman’s

niman?take

(AELS 24.188)

‘Why should we take those of another man?’c. To

tohwæmwhom

locigelook

icI

butonexcept

toto

ðæmthe

eaðmodum. . . ?humble

(CP 298,19)

‘To whom shall I look but to the humble. . . ?’

Second, with initial negation:

(7) a. nenot

mihtoncould

hithey

nænignenot-any

fultumhelp

ætfrom

himhim

begitanget

(Bede 1.10,15)

‘they could not get any help from him’b. Ne

notscealshall

hehe

nohtnothing

unalyfedesunlawful

dondo

(CP 60,14)

‘He must not do anything unlawful’

And third, with the adverbs þa and þonne in initial position:

(8) a. þathen

ge-mettemet

hehe

sceaðanrobbers

(AELS 31.151)

‘Then he met with robbers.’b. þa

thenforonsailed

hiethey

midwith

þrimthree

scipumships

utout

ongenagainst

hie,them

(ChronA 897)

‘then they sailed out with three ships against them’c. þonne

thentodælaðdivides

hihe

hishis

feoh,property

(Or 17.13)

‘then he will divide his property,. . . ’3According to van Kemenade and Westergaard (2011), some dialects of Norwegian permit a pronominal subject between a monosyllabic

wh-phrase and the finite verb. In terms of the analysis presented here (section 4), these dialects permit CP recursion above C* just in thisenvironment, and only with a fronted pronoun.

4

As can be seen from the above examples, pronouns in these three cases appear not before the finite verb, but afterit.4

3 Previous Analyses and Issues

Several analyses have been proposed for V3 and V4 orders in Old English. I discuss three here in broad terms, andshow how none of them are satisfactory.

3.1 The Clitic Analysis

One analysis proposes that V3 and V4 word order is not due to the V2 syntax of Old English, but to the peculiaritiesof pronouns. The idea is that pronouns are clitics, which means that their order is determined at least partially on anon-syntactic basis. Numerous publications refer to pronouns in Old English as clitics; some prominent examplesinclude van Kemenade (1987) and Kiparsky (1995). Both van Kemenade and Kiparsky propose analyses thatinvolve more than just this (Kiparsky 1995 in particular is very detailed and attempts to capture some of the sameobservations as this paper), but here I wish to focus only on the claim that pronouns are clitics.

The basic idea behind a clitic analysis is to say that pronominal clitics simply do not count for V2. The syntaxassembles a V2 clause, and pronouns cliticize somewhere at the front of the clause. The problem with this analysisis that V3 and V4 order can result from elements besides pronouns. Three examples are shown here, with a full NPsubject following a fronted non-subject:

(9) a. nunow

ealleall

ðasthese

ðingthings

sindare

midwith

anumone

namanname

genemnode,named

gesceaft.creature

(AECHom I 20,22)

‘Now all these things are called with one name: creature.’b. Sumum

somemonnumpersons

GodGod

seleðgives

ægðer geboth

goodgood

geand

yfelbad

gemenged,. . . (Bomixed

39.11,3)

‘God gives some people both good and bad things.’c. On

inGrecisreGreek

spræcelanguage

steorrastar

ysis

asteraster

genemned.called

(ByrM iii.3,189)

‘A star is called aster in the Greek language’

Koopman (1998) and Haeberli (2002a, 2002b) show that rates of V3 order with a fronted non-subject and a non-pronominal subject range from 6% to as high as 59.5% in different Old English texts. It is therefore not possible tomaintain that V3 is the result solely of the clitic status of pronouns.

This analysis also does not connect Old English word order to patterns in Present-Day English that I will showare strikingly similar (section 5). It also does not locate Old English within a typology of V2 languages, none ofwhich have V3 or V4 word order with pronouns (even though pronouns in many Germanic languages are oftenviewed as clitics). In addition, as far as I can tell, there is no independent reason to treat pronouns as clitics in OldEnglish. They do not uniformly appear in certain positions, nor do they gravitate toward certain elements like the

4There are also cases of verb-initial and verb-final main clauses, like the following:

(i) a. wæswas

sethe

frumabeginning

þusthus

awritenwritten

(Bede 1.10,10)

‘The beginning was written as follows:. . . ’b. Se

themanfullaevil

gastspirit

þathen

martineMartin

gehyrsumode.obeyed

(AELS 31.1050)

‘Then the evil spirit obeyed Martin.’

I will not be concerned with these here, and will just assume that they are stylistically marked variants that are allowed by the in-principleoptional processes of topicalization and verb-fronting. See Pintzuk 1999 for discussion. The remainder of this paper will be concerned withV2, V3, and V4 order.

5

finite verb. The analysis that I will propose treats them as different from other NPs only in their propensity to frontto clause-initial position.

3.2 The V-to-Infl Analysis

Versions of the V-to-Infl analysis have been proposed by Pintzuk (1993, 1999), Kroch and Taylor (1997), and others.The basic idea is that in Old English, the main verb moves not to C as in standard analyses of V2 (den Besten 1983),but to Infl. This was proposed as a way of accounting for embedded V2 in Icelandic and Yiddish (Rögnvaldssonand Thráinsson 1990, Diesing 1990, Santorini 1992). In the proposed typology, V-to-C V2 languages like Germanand Dutch do not allow embedded V2, because the C position is taken by the complementizer. If V2 is instead V-to-Infl, then V2 can still occur below a complementizer. In V-to-Infl languages, Spec-IP is not a dedicated subjectposition, but is instead a general topic position. The subject can remain in a low position within VP, while any XPcan move to Spec-IP. The finite verb moves to Infl. In Old English main clauses, if some other constituent alsomoves to Spec-CP or to some adjoined position, then the result will be V3 (XP in Spec-CP followed by YP inSpec-IP followed by V in Infl).

An attractive part of this account is that it attempts to locate Old English in a typology of V2 languages. It alsoseems to be supported by the fact that Old English does have V2 embedded beneath a complementizer (3), just likeIcelandic and Yiddish. A major problem, though, is that Icelandic and Yiddish never have V3 or V4 order in mainclauses. This discrepancy is completely unexpected on the V-to-Infl account, according to which Old English issupposed to be just like Icelandic and Yiddish.

Another issue is that this account does not relate Old English to patterns of word order in Present-Day English.As I will show in section 5, there is significant continuity between Old and Present-Day English in numerousrespects, which an adequate analysis should account for.

3.3 Two Subject Positions

Another analysis, proposed in various forms, says that Old English had two distinct subject positions (Fischer et al.2000; Haeberli 2000, 2002b, 2002a; van Kemenade 2012; van Kemenade and Westergaard 2011; Haeberli andIhsane 2016; among others). The left periphery of the clause includes CP with its specifier and head positions, thensome FP. Spec-FP is the position for pronominal subjects and subjects that are old information. F then selects TP,and Spec-TP is the position for subjects that are new information:

(10) [CP XP C [FP pronoun/old F [TP subject(new) . . . ]]]

Certain fronted XPs trigger V movement to C; these are wh-XPs, negation, and the adverbs þa and þonne fromabove. With these elements, the V will move to C and will precede both subject positions, resulting in strict V2.In other cases, however, the verb only moves to the head F. If an XP fronts to Spec-CP and the subject is oldinformation like a pronoun and therefore moves to Spec-FP, the order will be XP Subject V.

Like the previous analysis, this analysis does not relate Old English to Present-Day English in any appreciableway. It also does not locate Old English in a typology of V2 languages, unlike the V-to-Infl analysis. On theempirical side, the order Subject-XP-V is also attested as an instance of V3:

(11) a. icI

þæmthat

godangood.man

scealmust

forfor

hishis

modþræcedaring

madmastreasures

beodan.offer

(Beo 384–385)

‘I must offer treasures to that good man for his daring.’b. &

andicI

gehwameveryone

willewill

þærtothereto

tæcan. . . (Ordirect

57.15)

‘. . . and I will direct everyone there. . . ’c. þæt

thathehe

þætthe

godesGod’s

hushouse

woldewould

midwith

fyrefire

forbærnanburn

(AELS 25,613)

‘. . . that he would burn the house of God with fire. . . ’

6

This order is not permitted by the two-subject template, which only allows a subject to follow a fronted XP.Another empirical problem is that V3 often results from object pronouns as well as subject pronouns:

(12) a. GodGod

himthem

worhtewrought

þathen

reafgarments

ofof

fellumskins

(AECHom I 1,147)

‘then God made garments of skin for them’b. Fela

manyspellastories

himhim

sædontold

þathe

BeormasPermians

ægþer geboth

ofof

hieratheir

agnumown

lande,. . . (Orcountry

14,27)

‘the Permians told him many stories, both about their own country’

There is a variant of this analysis where the position for pronominal subjects is a position for pronouns in general,not just subject pronouns (Fischer et al. 2000). However, even this analysis runs afoul of examples like those in(11). There are also examples of V3 word order where neither of the XPs before the finite verb is a pronoun (or asubject), like the following:

(13) a. Onin

þæmthose

dagumdays

onin

TraciaThrace

þæmthe

londeland

wæronwere

twegentwo

cyningaskings

ymbabout

þætthat

ricekingdom

winnendefighting

‘In those days, in the land of Thrace, two kings were quarreling about that kingdom.’ (Or 63,7)b. Ðysne

thisyrmingpoor-wretch

æfterafter

hishis

forðsiðedeath

wurðodonworshiped

þathe

hæðenanheathens

eacalso

forinstead-of

healicnehigh

god,God

‘After his death, the heathens also worshiped this poor wretch instead of God.’ (WHom 12,60)

Koopman (1998) shows that multiple fronting of non-subjects is quite common (23 V3 examples in just one text,AECHom I). Additionally, there is a good number of exceptions to the dominant V2 pattern with negation, like thefollowing:

(14) a. onin

cristesChrist’s

namanname

nenot

forhtigefear

icI

forfor

þinumyour

tintregum;torments

(AECHom I 29,197)

‘In the name of Christ I fear not for your torments’b. buton

withouttwyndoubt

nenot

mihtecould

hehe

beonbe

ælmihtigAlmighty

godGod

(AECHom I 20,27)

‘without doubt he could not be Almighty God’

If the negative element ne is viewed as the first constituent in the dominant pattern in (7), so that the dominantpattern of V2 with negation is indeed V2 (and not V1), then such examples have V3 word order with two frontedelements that are neither pronouns nor subjects nor even arguments.

All the facts therefore indicate that V3 and V4 word orders in Old English are not the result of two subjectpositions, or a dedicated position for pronouns.

3.4 Summary

As can be seen, none of the analyses that have been proposed are satisfactory. None adequately locate Old Englishwithin a typology of V2 languages. None relate the Old English facts to Present-Day English. All struggle withvarious empirical facts.

In the next section, I will propose a simple typology of V2 languages with a small number of points of variation.Old English fits squarely within this typology. I will also show (in section 5) that Old English and Present-DayEnglish have many facts in common, which need to be related. The analysis that I propose does so relate them, andproposes complete historical continuity in many respects.

7

4 A Typology of V2

In this section I build a typology of V2 languages from the analysis of English subject-auxiliary inversion inBruening (2016). Bruening (2016) makes two proposals to account for ordering facts across languages in theleft periphery. The first has been proposed before, namely, that one CP can embed another in an instance ofCP recursion (Reinhart 1981, Platzack 1986, Bhatt and Yoon 1991, Culicover 1991, Authier 1992, Vikner 1995,McCloskey 2006, among others). The idea is that fronted elements are always in a specifier of CP and are neveradjoined. If there are multiple fronted elements, then there are recursive CPs. Similarly, if V2 can be embeddedbelow a complementizer, then that complementizer must be able to merge with another CP as its complement.To illustrate with English, negative inversion can occur below a complementizer, and this is an instance of CPrecursion, as shown in (15b):5

(15) a. She said that never again would she buy vinyl socks.b. . . . said [CP that [CP never again [C would [TP she . . .

An indication that CP recursion is the correct analysis is that multiple complementizers can sometimes bepronounced, one on each side of a fronted constituent (McCloskey 2006):

(16) a. He says that if that happens that he must be warned immediately.b. My fervent prayer is that for the sake of the president and the sake of this nation that this matter is

resolved soon. (AP wire report cited by McCloskey 2006, (69c))

CP recursion accounts for this better than a split CP with Force, Finiteness, and Topic and Focus projections (Rizzi1997); in such a split CP, it is not clear how two different positions could both be pronounced as the complementizer,which is supposed to be one (or more) of the dedicated heads. (See Abels 2012 for further criticism of the split CPapproach.)

The second proposal is that, in many languages, some C heads are distinguished as having certain properties,chief among them that they require movement of the finite verb to that C head. In Bruening (2016), this followsfrom a phonological Alignment theory (McCarthy and Prince 1993), where the tensed verb is required to Alignwith some projection of the distinguished C head. The Alignment aspect of this proposal will not be importanthere. What is important is that languages have such a dedicated C head, which Bruening (2016) designates “C*,”and this C* always requires movement of the finite verb to its position. We can think of C* as a feature assignedto a C head, or the class of C*s as a subset of the lexical items that are Cs, or we could define C* configurationallywithin a given language. Whatever approach we choose, the important part is that languages can choose which Cin cases of CP recursion is the distinguished C*.

In Present-Day English inversion contexts (questions, negative inversion, conditionals, etc.), the lowest C in CPrecursion is C*. This results in it being possible for elements to intervene between a fronted wh-phrase or negativephrase, say, and the fronted finite verb:

(17) a. And why in Paris did the Americans modify the agreement at the last minute. . . ? (The Guardian,cited by Haegeman 2000, note 2)

b. To whom at last will the government turn? (The Guardian, cited by Haegeman 2012, 51, note 49)c. I promise that on no account during the holidays will I write a paper. (Sobin 2003)d. What under no circumstances would John do for Mary? (Maekawa 2006, 230, (6a))

Elements can also come to the left of fronted wh-phrases and negative phrases:

(18) a. And at that point, who did he visit?b. Next Christmas whose parents should we go to? (McCloskey 2006, (32a))

(19) a. During the holidays, at no point did I write a paper.5I use T(ense)P for the highest functional head below C, but this is equivalent to IP as described above for the V-to-Infl analysis.

8

b. Most of the time, when she is working on a paper, only rarely does she leave her office. (McCloskey2006, (33a))

In the CP recursion analysis, these have CP recursion, with fronted elements in Spec-CPs and fronting of thefinite verb to the lowest C, which is the dedicated C*:

(20) a. [CP to whom [CP at last willC* [TP the government turn]]]?b. I promise [CP that [CP on no account [CP during the holidays willC* [TP I write a paper]]]].

Since the lowest C is C*, nothing can intervene between the fronted V and the subject in Spec-TP (Rizzi 1997,Haegeman 2012):

(21) a. When at last the sun came up,. . .b. * When will at last the sun come up?

(22) a. I think that, unfortunately, the gorilla has escaped.b. * Only once has, (un)fortunately, the gorilla escaped.

(23) (Rizzi 1997, (59))a. If yesterday John had done that, . . .b. * Had yesterday John done that,. . .

The (a) examples are embedded clauses, where there is no C*. In these clauses, elements can freely intervenebetween a higher C and the subject in Spec-TP:

(24) a. [CP if [CP yesterday [TP John had done that]]]. . .b. [CP when [CP at last [TP the sun came up]]]. . .

In the (b) examples, however, nothing can come between the finite verb in C* and the subject in Spec-TP. This isbecause C* is always the lowest C, and there is no such thing as adjunction to TP:6

(25) a. * [CP had [CP yesterday C* [TP John thad done that]]]. . .b. * [CP when will [CP at last C* [TP the sun come up]]]?

These are ungrammatical because the verb has moved to the higher C, which is not C* in English.One point of cross-linguistic variation in this system is now which C in CP recursion is the distinguished C*.

Strict V2 languages (all Germanic languages except English) designate the highest C as C*. This means that onlyone Spec-CP can precede the finite verb in C*, and we have strict V2. German is an example, where any XP canprecede the finite verb, but only one:

(26) a. Imin.the

Parkpark

spieltenplayed

diethe

Kinderchildren

vorbefore

derthe

Schuleschool

Fussball.soccer

b. vorbefore

derthe

Schuleschool

spieltenplayed

diethe

Kinderchildren

imin.the

Parkpark

Fussball.soccer

c. * Imin.the

Parkpark

vorbefore

derthe

Schuleschool

spieltenplayed

diethe

Kinderchildren

Fussball.soccer

d. * [CP Im Park C* [CP vor der Schule spielten [TP die Kinder Fussball ]]]

6In Bruening (2016), this adjacency requirement follows from the proposed Alignment theory. All that is important here is which C inCP recursion is C*.

9

In this system, strict V2 is the result of requiring the highest C in CP recursion to be C*. (26d) is ungrammaticalbecause the verb has failed to move to C*.

A second point of variation is whether CP recursion is allowed at all. Some V2 languages permit a frontedconstituent to come between the finite verb and the subject, while others do not. According to Haeberli (2000,112), German, (some speakers of) Dutch, Frisian, Yiddish, Swedish, and Norwegian permit a constituent in thisposition. The following example illustrates German:

(27) Wahrscheinlichprobably

wirdwill

späterlater

HansHans

dieselbethe.same

Uhrwatch

kaufen.buy

‘Hans will probably buy the same watch later.’ (Haeberli 2000, (6a))

Based on facts from particle placement in German, the subject in such cases is high, in Spec-TP. Particles like jaand doch, which are standardly taken to separate TP from VP (Webelhuth 1992, Diesing 1992, Struckmeier 2014,among others), follow the subject when it is preceded by an adverb (Solveig Bosse, Uli Sauerland, p.c.):

(28) Vielleichtmaybe

wirdwill

späterlater

HansHans

ja/dochPART

dieselbethe.same

Uhrwatch

kaufen.buy

‘Maybe Hans will buy the same watch later.’

This means that the adverb must be higher than TP; given that adjunction to TP is never allowed, it must be in aspecifier of a lower CP. I take this to indicate that German permits CP recursion below C*, as follows:

(29) CP

XP

wahrscheinlich

C

C*wird

CP

YP

später

C

C TP

NP

Hans

T

dieselbe Uhr kaufen

In contrast, as described above, Present-Day English permits CP recursion only above C*:

(30) CP

XP

to whom

C

C CP

YP

at last

C

C*will

TP

NP

the government

T

turn

10

Some other Germanic languages, namely, W. Flemish, Afrikaans, Icelandic, and Danish, do not allow anythingto come between the finite verb and the subject (Haeberli 2000). The following example is Icelandic:

(31) Sennilegaprobably

munwill

(*seinna)(*later)

JónJohn

kaupabuy

samathe.same

úrið.watch

‘Probably, John will buy the same watch later.’ (Haeberli 2000, (6g))

This means in the CP recursion analysis that these languages do not permit CP recursion at all (in main clauses).Only one constituent can precede the finite verb, and nothing can follow it before the TP boundary.

A third point of variation concerns whether some embedded Cs can take a C*P as their complement. Icelandicand Yiddish, which allow embedded V2 below a complementizer, do allow this, as the following Yiddish exampleshows:

(32) Iryou.Pl

zoltshould

visnknow

. . . , azthat

vaynwine

kencan

menone

makhnmake

funfrom

troybngrapes

oykh.also

‘You should know that one can make wine from grapes also.’ (Diesing 1990, 44, (5a))

Present-Day English also allows this, as in example (17c) above.These three points of variation lead to the following typology of V2 languages:

(33) Main Clauses: CP Recursion?a. No: W. Flemish, Afrikaans, Icelandic, Danishb. Yes, C* is Highest: German, Dutch, Frisian, Yiddish, Swedish, Norwegianc. Yes, C* is Lowest: Present-Day English

(34) Embedded Clauses: CP Recursion?a. Yes: Icelandic, Yiddish, Present-Day Englishb. No: German, Dutch, Swedish, Norwegian, Danish, W. Flemish, Afrikaans

(Additionally, Present-Day English only has C* in a restricted set of environments. This is a fourth point ofvariation: whether C* occurs in all matrix clauses, or only in restricted environments.)

We are now in a position to see how Old English fits into this typology, in the next section.

5 Old English from the CP Recursion Perspective

We can now examine Old English from the perspective of the typology outlined in the previous section. As wewill see, Old English is like Present-Day English in permitting recursion above C*, but it is also like German inpermitting CP recursion below C*.

5.1 Recursion Above C*

We have already seen that Old English permits more than one constituent to precede the finite verb in a V2 context,exactly like Present-Day English and unlike all modern V2 languages. Two examples of more than one constituentbefore the finite verb are repeated below, with an analysis in terms in CP recursion (strikethrough indicates anunpronounced copy, in a lower argument position):

(35) a. hehe

hinehim

eftagain

ongonbegan

wætereswater

weorpan,sprinkle

(Beo 2790–2791)

‘He began to sprinkle him again with water. . . ’b. [CP he [CP hine [CP eft ongon [TP he hine wæteres weorpan ]]]]

11

(36) a. &and

icI

gehwameveryone

willewill

þærtothereto

tæcan. . . (Ordirect

57.15)

‘. . . and I will direct everyone there. . . ’b. [CP ic [CP gehwam wille [TP ic þærto tæcan]]]

Like Present-Day English, C* must be the lowest C in CP recursion environments in such cases.This is exactly a point of historical continuity that I would like to emphasize here, since it unifies Old and

Present-Day English and distinguishes both of them from all other Germanic languages. It is striking that Present-Day English regularly puts multiple constituents at the beginning of the sentence:

(37) a. At the same time, back on the ranch, the cowhands were readying their firearms.b. Then, before anyone could react, without even thinking about it Susan ran for the exit.

This is very different from all other Germanic languages, which strictly limit the constituents at the beginning ofthe clause to one before the finite verb. This is what makes them V2 languages. In Present-Day English, though,even in inversion contexts (the counterpart of V2), multiple constituents can precede the finite verb, as was shownabove. Additional examples follow:

(38) a. Why at that point did you not call the police?b. After the liquid starts boiling at no point should you allow the other chemical to come into contact

with it.

In this Old English patterns with Present-Day English. Pintzuk (1999, 65–68) and Kroch and Taylor (1997, sec.3.4)note that Old English main clauses often have fronted phrases before the initial constituent, as in the followingexamples:

(39) a. &and

onin

þamthe

.xlii.42nd

geareyear

hishis

ricesreign

CristChrist

wæswas

acenned.born

(ChronA 1)

‘And Christ was born in the 42nd year of his reign.’b. þæs

thatcyningesking’s

tidumtime

sethe

ArrianiscaArian

gedwolaheresy

wæswas

upcumenarisen

(Bede 1.8:16)

‘The Arian heresy arose in that king’s time.’c. On

inþisumthis

geareyear

WillelmWilliam

cyngking

geafgave

RaulfeRalph

eorleearl

WillelmesWilliam’s

dohtordaughter

OsbearnesOsborn’s

sunu.son

(ChronE

1075)

‘In this year King William gave William FitzOsborn’s daughter in marriage to Earl Ralph.’d. &

andfullicefully

.lxx.70

wintrayears

syððanafterwards

on ancontinually

wæswas

sethe

ðeodscypenation

eallall

geðeowodenslaved

underunder

heoratheir

feondaenemies’

gewealde,power

(WHom 6.120)

‘And for fully 70 years afterwards, all the nation was continually enslaved under their enemies’power. . . ’

e. þathen

æfterafter

sumumsome

fyrstetime

hehe

wearðwas

onin

swefnedream

gemynegodadmonished

(AELS 31,145)

‘Then after some time he was admonished in a dream’

Examples like these show that V3 and V4 orders are not just the result of pronouns acting as clitics, or theresult of two subject positions. In (39d), for instance, the verb is third, but the two constituents preceding it areboth adverbial expressions. In (39e), the verb is fourth, but two of the three constituents preceding it are adverbialexpressions and only one is a subject pronoun.

12

Even in the contexts that are supposed to require strict V2 like negation and with the adverb þa in initial position,it is possible for constituents to come to the left of the initial constituent. Some examples of this with negationappeared above in (14). Similarly, in (40a), a left-dislocated NP precedes negation, which as initial constituentis supposed to strictly require V2. Note that the pronoun subject follows the verb in this case, as is typical withnegation. This might be a true instance of left-dislocation, where the dislocated element does not even count forV2, but even this is striking because it is not allowed in the modern Germanic languages. According to Kroch andTaylor (1997), a left-dislocated NP can only precede the initial constituent of a V2 sentence in German if the initialconstituent is the pronoun that is coindexed with it. This is not the case in (40a), where the initial constituent is thenegative particle ne.

(40) a. Ðathe

onsaegdnysse,sacrifices

þa ðewhich

framby

eowyou

deoflumdevils

wæronwere

agoldene,offered

nenot

magoncan

hithey

ðamthe

underðeoddumdevotees

gefulltumianhelp

(Bede 1.7,49)

‘The sacrifices, which you offered to devils, they can’t help the devotees. . . ’b. þa

thenunder þæmmeanwhile

þathen

bestælstole

hehe

hinehim

onin

nihtnight

onwegaway

‘Then, meanwhile, he stole away in the night. . . ’ (ChonA 901)c. Mid

withþythat

ðathen

ongonbegan

firenlustriotous-living

weaxanincrease

(Bede 1.11,15)

‘With that, riotous living then began to increase.’

Similarly, in (40b), the verb is fourth, following the adverb þa. The subject pronoun follows the finite verb. At thesame time, however, two other constituents precede þa, neither of which is a pronoun or a subject. In (40c), theverb is third, following a PP and the adverb þa. V3/V4 word order therefore has nothing to do with clitic pronounsor subject positions.

I conclude from this that Old English was just like Present-Day English in fronting constituents to specifiersof multiple CPs above C* in CP recursion. Both Old English and Present-Day English differ significantly fromall other modern Germanic languages in this respect. All other Germanic languages allow only one constituentbefore the finite verb.7 (According to Kroch and Taylor 1997, note 8, older West Germanic languages toleratedconstituents on the left to a much greater degree than do the modern Germanic languages. If this is correct, thenEnglish may preserve the older pattern, and strict V2 is a more recent innovation.)

In terms of the typology outlined in the previous section, then, Old English was just like Present-Day Englishin permitting CP recursion in main clauses. It was also just like Present-Day English in having the lowest C in suchcases be C*. The one difference that we have to recognize so far is that Old English, unlike Present-Day English,regularly fronted pronouns to the left edge of the clause, including to Spec-CP in a recursive CP. One example ofthis is repeated below:

(41) hehe

hinehim

eftagain

ongonbegan

wætereswater

weorpan,sprinkle

(Beo 2790–2791)

‘He began to sprinkle him again with water. . . ’

Pronouns often move to Spec-CP, and in examples like the above, this puts them to the left of the finite verb, whichmoves to the lowest C (C*).

This fronting of pronouns to the left periphery can be seen independently in subordinate clauses where there isno C* and no V2:

7According to McCloskey (2006, note 6), while most Germanic languages strictly ban more than one constituent before the finite verbin declarative clauses, many do permit a constituent to occur in an interrogative clause, before the wh-phrase. Apparently these languagesdo permit a restricted form of CP recursion just in this context. See also note 3.

13

(42) a. &and

icI

forþamtherefore

hitit

nunow

yldandelay

nenot

mægmay

(GD(H) 21,19)

‘. . . and therefore I may not delay it now.’b. Gif

ifhimthem

ðonnethen

GodGod

ryhtlicerightly

&and

strecliceseverely

demanjudge

wilewill

(CP 44.21)

‘If God will then judge them rightly and severely. . . ’c. &

andheomthem

manone

syððanafterwards

þærthere

friðpeace

wiðwith

nammade

(ChronA 1001))

‘And afterwards they made peace with them there.’

Fronting of pronouns takes place even in the three contexts where strict V2 predominates. In these contexts thepronoun occurs to the right of the finite verb, but it can be seen clearly to have fronted, for instance when it is anobject (43a–c) or when it strands a preposition (43d):

(43) a. Nenot

geseahsaw

hinehim

nanno

manman

nates-hwonso-little

yrreangry

(AELS 31,306)

‘No man ever saw him so little angry.’b. þa

thensticodestuck

himhim

monsomeone

þathe

eaganeyes

ut,out

(Or 90.14)

‘then his eyes were gouged out,’c. þonne

thenmotmay

hinehim

sethe

hlafordmaster

gefreogeanliberate

(LawIne 74,1)

‘Then the master may liberate him.’d. þa

thenbecomcame

himhim

AntigonesAntigones

midwith

firdearmy

onagainst

(Or 79.23)

‘. . . then Antigones came against him with an army. . . ’

Pronouns can also be seen to have moved when there is a doubling negative particle after the verb. Fischer et al.(2000, 125) observe that pronouns occur to the left of the doubling negative particle (na in the examples below),but full NPs occur to the right:

(44) a. Nenot

hetordered

hehe

usus

naNeg

leornianlearn

heofonasheavens

toto

wyrcennemake

(AELS 16,127)

‘He did not bid us learn to make the heavens’b. Ne

notwendethought

naNeg

EzechiasEzechias

IsrahelaIsrael’s

kyningking

ðætthat

hehe

gesyngade. . . (CPsinned

39.2)

‘Ezechias, king of Israel, did not think he was sinning. . . ’

In the CP recursion theory, we can account for these facts by moving pronouns to specifiers of recursive CPs,as shown below:

(45) a. hehe

hinehim

eftagain

ongonbegan

wætereswater

weorpan,sprinkle

(Beo 2790–2791)

‘He began to sprinkle him again with water. . . ’b. [CP he [CP hine [CP eft ongon [TP he hine wæteres weorpan ]]]]

Fronting to specifiers of recursive CPs can put pronouns first, or second, or even third:

(46) a. &and

icI

gehwameveryone

willewill

þærtothereto

tæcan. . . (Ordirect

57.15)

‘. . . and I will direct everyone there. . . ’

14

b. Ælceach

yfelevil

hehe

mægcan

don,do

(WHom, 4.62)

‘He can do each evil.’c. &

andseofonseven

ærendracanmessengers

hehe

himhim

hæfdehad

toto

asend.send

(ChronA 905)

‘He had to send seven messengers to him.’

In other words, there is no dedicated position for pronouns, they simply need to front to some Spec-CP. Additionally,both subject and object pronouns regularly front, although fronting is more common with subject pronouns (e.g.,van Kemenade 1987).

This analysis therefore accounts for the pronoun data much better than the analysis that proposes two subjectpositions. Pronouns clearly do not occupy fixed positions, and the phenomenon is not limited to subjects.

5.2 Recursion Below C*

In many cases, then, Old English has the lowest C in CP recursion be C*, just like Present-Day English. However,wh-phrases, negation, and the adverbs þa and þonne generally do not have recursion above the finite verb. Instead,they seem to have recursion below the finite verb. As we just saw, fronted pronouns are clearly fronted, and yetthey are often to the right of these elements (and the finite verb). Apparently, with wh-phrases, negation, and theadverbs þa and þonne, C* is usually the highest C in CP recursion rather than the lowest. When a pronoun fronts,it fronts to a specifier of a CP below C*, as shown in (47b):

(47) a. þathen

becomcame

himhim

AntigonesAntigones

midwith

firdearmy

onagainst

(Or 79.23)

‘. . . then Antigones came against him with an army. . . ’b. [CP þa becomC* [CP him [TP Antigones mid firde on him]]]

In other words, just with wh-phrases, negation, and the adverbs þa and þonne, Old English is just like other Ger-manic languages in C* being the highest C in CP recursion. Even this is not completely strict, however, givenexamples like (40a–40b) and (14).

Other constituents can also occupy Spec-CPs below C*. As Haeberli (2000) shows, Old English is just likeGerman in permitting a constituent between the finite verb in a V2 clause and the subject:

(48) a. Nenot

dorstedared

swa þeahhowever

sethe

mæssepreostmass-priest

þonethe

bisceopbishop

geaxianask

for hwan. . .why

‘However, the priest did not dare to ask the bishop why. . . ’ (GD(H) 58,3)b. þa

thenwæswas

inin

þathat

tidtime

UitaliusVitalius

papapope

þæsthe

apostolicanapostolic

seðlessee’s

aldorbiscophigh-priest

‘At that time, Vitalius was chief bishop of the apostolic see.’ (Bede 4.1,15)c. In

inðathat

tidtime

wæswas

inin

MercnaMercians’

mægðecountry

WulfhereWulfhere

cyning.king

(Bede 4.3,1)

‘At that time, Wulfhere was king in Mercia.’

It therefore appears that Old English had two grammars. One has been maintained to Present-Day English.In this grammar, the lowest C in CP recursion is C*. This results in V3 or even V4 word order with multiplefronting to Spec-CPs. In the other grammar, the highest C is C*. This is the same grammar as German, Dutch,Frisian, Yiddish, Swedish, and Norwegian. In this grammar, the verb is always second in V2 clauses but frontedconstituents can come between the finite verb and the subject. With wh-phrases, initial negation, and initial þa andþonne, this grammar predominates. It it not limited to these contexts, however, as (48c) illustrates.

15

5.3 The Relation between the Two Grammars

The question that arises now is what the relation between these two grammars is. One possibility is that they wouldbe in competition, and so any given Old English clause would have either recursion above C*, or below C*, butnot both. To test this, I used the CorpusSearch program to search the York-Toronto-Helsinki Parsed Corpus of OldEnglish Prose. I searched for cases where the finite verb precedes the subject but something else comes betweenthem. I further narrowed this down to cases where the subject precedes the non-finite verb, because without thisrestriction, numerous cases turn up where the subject is at the end of the clause, typically as a negative or indefinite.I assume that in these types of clauses, the negative or indefinite subject is not in the canonical subject position,Spec-TP, and so cannot tell us whether recursion is taking place above or below C* (the position of the finite verb).

461 examples turned up in this search. The majority had only one constituent before the finite verb, or none atall (they were verb-initial). However, there were several classes of cases with more than one constituent before thefinite verb and one or more constituent between the finite verb and the subject. One such class can be dismissedimmediately. This is the case of if-clauses. It is common for an if-clause to be followed by þonne and then thefinite verb. In such cases, other elements can come between the finite verb and the subject. However, it appears thatif-clauses simply do not count for V2. They are left-dislocated, and resumed by the element þonne, which seems tobe playing a pronominal role in this use. I therefore exclude these from consideration, as they do not have recursionabove C*.

Even dismissing if-clauses, there were still several classes of cases with multiple elements to the left of thefinite verb and at least one constituent between the finite verb and the subject. The largest such class involvesunaccusative and (predominantly) passive main verbs. I give one example of each below:

(49) a. þathen

nenot

mihtecould

swaðeahhowever

seothe

RomanaRoman

burhtown

butonwithout

papanpope

wunian.continue

(AECHom II, 9:75.93)

‘But the city of Rome might not, however, continue without a pope’b. þy

thesyxtansixth

dægeday

ærbefore

undernethe.third.hour

þonnethen

biþbe

fromfrom

feowerfour

endumends

þæreof.the

eorþanearth

eallall

middangeardworld

midwith

awergdumaccursed

gastumspirits

gefylled,filled

(HomS 7:93,35)

‘On the sixth day before the third hour from the four ends of the earth all the world shall then befilled with accursed spirits,’

The nominative subject in these types of clauses, which is underlyingly an object, seems to remain in a low position,perhaps even in object position (cf. Haeberli 2000, 116–117). If this is indeed what is happening, then these typesof clauses also do not tell us whether recursion is above C* or below C*. (Given that they have more than oneconstituent before the finite verb, by assumption they have recursion above C*.) It should also be noted that inmany of the passive cases that appeared in this search, a dative NP or PP immediately followed the finite verb, asin example (49b). It is possible that this NP/PP occupies the subject position (Spec-TP).

The next class of examples with more than one constituent before the finite verb and a constituent betweenthe finite verb and the subject involved negative subjects. Negative subjects frequently appear to the right ofconstituents other than the finite verb. Here are some examples:

(50) a. þathen

nenot

meahtecould

hinehim

monno.one

gebindan;bind

(Bede 4.23,21)

‘Then no one could bind him’b. þa

thennenot

mehtecould

hinehim

monno.one

ofridan;overtake.riding

(ChronA 901)

‘No one could overtake him’c. þa

thennenot

mihtecould

naNeg

lengclong

mannaof.men

ænigany

hinehis

sylfneself

bedyrnan,conceal

(LS 23 [SevenSleepers], 114)

‘Then none of the men could any longer conceal himself,’

16

As mentioned above, negative and indefinite subjects frequently occupy a position other than Spec-TP, often atthe right edge of the clause. One possibility is that in examples like these, the subject is not in Spec-TP, and so itdoes not tell us where CP recursion is taking place. Since more than one constituent precedes the finite verb, byassumption recursion is taking place above C*. One possible location for the subject is its base position at the edgeof VP. In this analysis, pronouns that are fronted would not necessarily front to Spec-CP; they might, for instance,front to Spec-TP if the subject does not move there. Alternatively, these examples show us that recursion is possibleboth above and below C* at the same time. However, given that negative and indefinite subjects can be seen tooccupy other positions in general, it seems more likely that that is what is going on here.

So far, then, there is little reason to think that recursion can take place both above C* and below C* in thesame clause. However, there was a small number of examples (namely, five) that did not fall into one of the aboveclasses. I reproduce all of them here:

(51) a. þonethe.Acc

hehnelofty

GodGod

nenot

mæigcan

hinehim.Acc

manman

þurhby

ofermetapride

geræcen,reach

(Alc 35.98,7)

‘Man cannot reach lofty God through pride.’b. For þan

thereforenenot

scealshall

næfrenever

sethe

cristenachristian

manman

beonbe

orsorhleas.unsorrowless

(HomU 22.17)

‘Therefore the Christian man shall never be unsorrowless.’c. Forðon

whereforehinehim

mægmay

nunow

ælcevery

monman

oforswiþan,overcome

(HomS 3:31.98)

‘Wherefore now may every man overcome him;’d. Ac

andforhwywhy

neNeg

mægcan

ðonnethen

miclemuch

mamore

ðæsthe

modesmind’s

læcephysician

gehælanheal

ðathe

adlediseases

ðæraof.the

unðeawavices

monigraof.many

monnamen

midwith

anresame

lare,teaching

(CP 457,1)

‘Why cannot then much more the physician of the mind heal the diseases of the vices of many menwith the same instruction’

e. þythat

dægeday

nenot

mægmay

þeyou.Acc

sethe

gemædlachatter

sceþþan.injure

(Lch II 3:57.1)

‘that day the chatter cannot harm thee.’

Example (51a) appears to have a left-dislocated NP resumed by a pronoun after the finite verb. In such a case, theleft-dislocated NP might not count at all for V2, so that (51a) has only recursion below C*. The other four examplesappear to be genuine instances of recursion both above and below the finite verb. However, given how few suchexamples are, I tentatively conclude that it is reasonable to view the two grammars as being in competition. Aspeaker would either select the C*-highest grammar (giving strict V2 and the possibility of constituents betweenthe finite verb and the subject) or the C*-lowest grammar (with V3 or V4 and no constituent between the finite verband the subject). With wh-questions in particular the C*-highest grammar dominated, and to a lesser extent it alsodid with negation and the adverbs þa and þonne.

Consistent with this, many of the examples with a constituent between the finite verb and the subject are exactlythese three contexts. Many are wh-questions:

(52) a. Hwætwhat

mægcould

ðonnethen

elleselse

seothe

earcark

tacniansignify

butonexcept

ðathe

halganholy

ciricean,church

(CP 170,2)

‘What signifies the ark but the holy church,’b. Hu

howgeradesexactly

mægcan

ðonnethen

sethe

biscepbishop

brucanenjoy

ðærethe

hirdelicanpastoral

are,dignity

(CP 133,3)

‘How, then, can the bishop properly enjoy the pastoral dignity,’

Many have initial negation:

17

(53) a. Nenot

magoncan

þisthis

þeahhowever

ealleall

menmen

don;do

(HomS 3:37,34)

‘All men however cannot do this;’b. Ne

notmagancould

þonnethen

haligeholy

menmen

onin

þamthose

timantimes

ænigeany

tacnamiracles

openliceopenly

wyrcan,work

(WHom 5,62)

‘In those times holy men could not openly work miracles,’

And many have the adverbs þa and þonne in initial position:

(54) a. þathen

onganbegan

eftagain

þætthe

wifwoman

sprecanspeak

toto

þamthe

ealdanold.man

(LS 23,248)

‘Then began the woman again to speak to the old man,’b. þonne

thenwæswas

ongeanby

ðyssumthis

wæterscipewater.patch

glæsenglass

fætvessel

onon

seolfrenresilver

racenteagechain

ahangenhanging

‘By this piece of water was a glass vessel, hung on a silver chain’ (BlHom 17:209,4)

It appears, then, that recursion below C* predominantly took place in these three environments, as expected.

5.4 Old English’s Place in the Typology

We can now locate Old English within the typology proposed in section 4. Old English has two settings for mainclauses, one with CP recursion above C*, and one with CP recursion below C*. The one with C* highest is regularlyused with wh-questions, negation, and certain adverbs. In addition, Old English permits CP recursion below someembedded complementizers:

(55) Main Clauses: CP Recursion?a. No: W. Flemish, Afrikaans, Icelandic, Danishb. Yes, C* is Highest: German Dutch, Frisian, Yiddish, Swedish, Norwegian, Old Englishc. Yes, C* is Lowest: Present-Day English, Old English

(56) Embedded Clauses: CP Recursion?a. Yes: Icelandic, Yiddish, Present-Day English, Old Englishb. No: German, Dutch, Swedish, Norwegian, Danish, W. Flemish, Afrikaans

Note that two of these three settings are exactly the same as Present-Day English. These two continuities are listedbelow, along with a third pattern of the language identified above:

(57) Continuities with Present-Day English:a. Embedded clauses permit CP recursion.b. C* is lowest C in CP recursion.c. Multiple constituents regularly front to Spec-CP.

The second two distinguish Old English and Present-Day English from all other Germanic languages. These com-monalities have not been identified before, although they are striking, and strikingly different from other Germaniclanguages. The present analysis captures them, while no other analysis does.

There are also several ways in which Old English differed from Present-Day English:

(58) Differences from Present-Day English:a. The highest C can be C* (like modern V2 languages).b. Pronouns regularly front to Spec-CP.

18

Pronouns stopped fronting to Spec-CP. Interestingly, this change resulted in more V2 with pronoun subjects throughthe Middle English period, even while V2 was in general declining. As Haeberli (2002b), van Kemenade and West-ergaard (2011), and van Kemenade (2012) note, subject-verb inversion actually increased with pronoun subjectsduring the Middle English period, although subject-verb inversion overall was declining in frequency. The expla-nation for this is that pronouns, and subject pronouns in particular, stopped fronting to Spec-CP. When the verbmoved to C, it therefore inverted more with pronominal subjects, even though verbs were moving to C less and less(see more on changes with pronouns immediately below).

To the two differences listed above we can add the following, although they have not yet been a focus of thepresent paper:

(59) Further Differences from Present-Day English:a. Main verbs could move to C*.b. Most main clauses had C*.

These both changed to get to the present-day grammar. C* became restricted in its distribution to questions,conditionals, clauses with a preposed negative phrase (although this seems to have been a new innovation), and afew other environments. Additionally, main verbs lost the ability to move to C*, so that verb movement becamerestricted to auxiliary verbs.

Finally, the option of the highest C in CP recursion being C* was lost, so that in Present-Day English only thelowest C is ever C*.

5.5 Proposed Route of Change

I propose that these changes took place in the following manner:First, the general headedness of the language changed. The VP changed from OV to VO, and the TP changed

from VP-T to T-VP. This was already taking place in the Old English period, and has been argued to have beenlargely completed by around the year 1200 (van Kemenade 1987, citing Canale 1978 and Hiltunen 1983). Oneimportant consequence of this change is that it makes it harder for a language learner to distinguish movement toC from movement to T when it is the subject that precedes the finite verb (by far the most common pattern).

Second, pronouns stopped fronting to Spec-CP. In particular, object pronouns stopped fronting and largelybecame confined to post-verbal position. According to van Kemenade (1987, 200), object pronouns had stoppedfronting entirely by the mid-fourteenth century, and predominantly appeared after the verb as early as the twelfthcentury. In contrast, subject pronouns continued to occur before the finite verb for a longer period (van Kemenade1987, 197). That is, V3 word order predominantly occurred only with the subject as the second constituent (XP-Subj-V). I suggest that this was reinterpreted by language learners as the verb only moving as far as T. The positionsfor pronouns had already changed, object pronouns in particular, so rather than taking XP-Subject-V order toindicate fronting of the subject, language learners instead took it to indicate failure of the verb to move to C.Similarly for the V2 order Subject-V: rather than fronting of both the subject and the verb, it was reinterpreted asthe subject being in Spec-TP and the verb in T.

At the same time, however, V2 was still very robust with wh-questions. So the actual change was for C* tobecome restricted in its distribution, not to be lost altogether. Alongside the old grammar where almost all mainclauses had a C*, there now arose a grammar where C* only occurred in certain main clauses, chief among themwh-questions. When C* did not occur, the verb moved only as far as T. According to van Kemenade (1987, 183ff),V2 began declining around the year 1400. In the current account, that is when the new grammar arose where C*occurs only in a restricted set of main clauses. This grammar existed alongside the earlier one where C* occurredin all main clauses and gradually supplanted it. This took a long time, however, with V2 occurring well into theEarly Modern period.

One possibility is that the grammar where C* was the highest C in CP recursion, which was already tied towh-questions, negation, and certain adverbs, turned into a grammar where C* is restricted to certain environments.That is, because there was already a relation between a restricted set of environments and the C*-highest grammar,that grammar turned into one where C* is restricted in its distribution. At the same time, though, there was regular

19

fronting of more than one constituent to a preverbal position, so this grammar also turned into C* being lowest.That is, C* being lowest was generalized, while the C*-highest grammar turned into a C*-lowest grammar whereC* is restricted to questions and a few other environments.

The only other change that we need to get to the Present-Day English grammar is the restriction on verbmovement to auxiliary verbs. This restriction is independent of the development of V2, so I will not address it here.

5.6 Further Support: Two Dialects in Middle English

The present analysis receives further support from the two dialects of Middle English discussed by Kroch andTaylor (1997). According to Kroch and Taylor (1997), the southern dialects of Middle English are just like OldEnglish with respect to V2, at least initially. However, a northern dialect strikingly exhibits strict V2, just likemodern V2 languages. Kroch and Taylor (1997) suggest that this is due to influence from Scandinavians settlingin England at the time. However, this makes no sense in the V-to-Infl analysis that they adopt (described in section3.2). In their analysis, Old English V2 was of the V-to-Infl type rather than the V-to-C type. The Scandinavianof the time was probably also V-to-Infl, like modern Icelandic. It is not clear how the influence of one V-to-Infllanguage could turn another V-to-Infl language into a V-to-C language.

In the analysis proposed here, in contrast, Icelandic (and, presumably, earlier Scandinavian) is like all otherV2 languages in having the highest C be C* in main clauses. This influence pushed the northern Middle Englishdialect to generalize the option of the higher C being C*, so that the lower C never was. The result is strict V2.

Furthermore, Haeberli (2000) shows that the northern Middle English dialect also did not allow the order V2XP Subj, just like Icelandic. In current terms, this means the northern Middle English dialect did not allow CPrecursion below C*, either. Just like Icelandic, northern Middle English did not allow CP recursion at all in mainclauses.

In short, in the current typology it makes perfect sense for Scandinavian influence to push a Middle Englishdialect into strict V2. This dialect of Middle English took on the settings of modern Icelandic. One of these settingswas already an option in the language, so this was not a large change at all.

6 Conclusion

Old English V2, V3, and V4 receives a natural account in the typology of the left periphery proposed in Bruening(2016). In many ways Old English was just like Present-Day English, with CP recursion above C* and fronting ofmultiple elements. This results in more than one constituent being possible before the finite verb. This is a featurethat has been maintained from Old English to Present-Day English, but it is something that is not allowed in allthe other Germanic languages. In these languages, if CP recursion is allowed at all, it is only allowed below C*,so that fronted elements intervene between the finite verb and the subject in Spec-TP. Old English actually allowedthis option, too, in certain environments.

The proposed analysis not only locates Old English within a typology of V2, it also explains historical continu-ity from Old English to Present-Day English, while positing minimal change to get from Old English to Present-DayEnglish. The empirical coverage of the analysis is superior to other accounts that have been proposed, which allstumble on various aspects of the data. None of them connect the Old English facts to patterns of Present-Day En-glish. Finally, to the extent that the analysis is successful, it provides further support for the CP recursion analysisproposed in Bruening (2016).

ReferencesAbels, Klaus (2012), “The Italian Left Periphery: A View from Locality.” Linguistic Inquiry 43: 229–254.Authier, J.-Marc (1992), “Iterated CPs and Embedded Topicalization.” Linguistic Inquiry 23: 329–336.den Besten, Hans (1983), “On the Interaction of Root Transformations and Lexical Deletive Rules.” In Werner Abraham,

ed., On the Formal Syntax of Westgermania, Amsterdam: John Benjamins, pp. 47–131.

20

Bhatt, Rakesh, and James Yoon (1991), “On the Composition of COMP and Parameters of V2.” In Dawn Bates, ed., TheProceedings of the Tenth West Coast Conference on Formal Linguistics, Stanford: CSLI Publications, pp. 41–52.

Bruening, Benjamin (2016), “Alignment in Syntax: Quotative Inversion in English.” Syntax 19: 111–155.Culicover, Peter W. (1991), “Topicalization, Inversion and Complementizers in English.” In Denis Delfitto, Martin Ev-

eraert, Arnold Evers, and Frits Stuurman, eds., Going Romance and Beyond, Utrecht: University of Utrecht, pp.1–45.

Diesing, Molly (1990), “Verb Movement and the Subject Position in Yiddish.” Natural Language and Linguistic Theory8: 41–79.

Diesing, Molly (1992), Indefinites. Cambridge, MA: MIT Press.Fischer, Olga, Ans van Kemenade, Willem Koopman, and Wim van der Wurff (2000), The Syntax of Early English.

Cambridge: Cambridge University Press.Haeberli, Eric (2000), “Adjuncts and the Syntax of Subjects in Old and Middle English.” In Susan Pintzuk, George

Tsoulas, and Anthony Warner, eds., Diachronic Syntax: Models and Mechanisms, Oxford: Oxford University Press,pp. 109–131.

Haeberli, Eric (2002a), “Inflectional Morphology and the Loss of Verb-Second in English.” In David Lightfoot, ed.,Syntactic Effects of Morphological Change, Oxford: Oxford University Press, pp. 88–106.

Haeberli, Eric (2002b), “Observations on the Loss of Verb Second in the History of English.” In Jan-Wouter Zwart andWerner Abraham, eds., Studies in Comparative Germanic Syntax, Amsterdam: John Benjamins, pp. 245–272.

Haeberli, Eric, and Tabea Ihsane (2016), “Revisiting the Loss of Verb Movement in the History of English.” NaturalLanguage and Linguistic Theory 34: 497–542.

Haegeman, Liliane (2000), “Inversion, Non-Adjacent Inversion, and Adjuncts in CP.” Transactions of the PhilologicalSociety 98: 121–160.

Haegeman, Liliane (2012), Adverbial Clauses, Main Clause Phenomena, and the Composition of the Left Periphery.Oxford: Oxford University Press.

Healey, Antonette diPaolo, and Richard Venezky (1980), A Microfiche Concordance to Old English. Toronto: PontificalInstitute of Medieval Studies.

van Kemenade, Ans (1987), Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris.van Kemenade, Ans (2012), “Rethinking the Loss of Verb Second.” In Terttu Nevalainen and Elizabeth Closs Traugott,

eds., The Oxford Handbook of the History of English, Oxford: Oxford University Press, pp. 823–834.van Kemenade, Ans, and Marit Westergaard (2011), “Syntax and Information Structure: Verb-Second Variation in

Middle English.” In Anneli Meurman-Solin, Marııa José López-Couso, and Bettelou Los, eds., Information Structureand Syntactic Change in the History of English, Oxford: Oxford University Press, pp. 87–118.

Kiparsky, Paul (1995), “Indo-European Origins of Germanic Syntax.” In Adrian Battye and Ian Roberts, eds., ClauseStructure and Language Change, Oxford: Oxford University Press, pp. 140–169.

Koopman, Willem (1998), “Inversion after Single and Multiple Topics in Old English.” In Jacek Fisiak and MarcinKrygier, eds., Advances in English Historical Linguistics, Berlin: Mouton de Gruyter, pp. 135–150.

Kroch, Anthony, and Ann Taylor (1997), “Verb Movement in Old and Middle English: Dialect Variation and LanguageContact.” In Ans van Kemenade and Nigel Vincent, eds., Parameters of Morphosyntactic Change, Cambridge:Cambridge University Press, pp. 297–325.

Maekawa, Takafumi (2006), “Configurational and Linearization-Based Approaches to Negative Inversion.” InO. Bonami and P. Cabredo Hofherr, eds., Empirical Issues in Syntax and Semantics 6, Electronic publication:http://www.cssp.cnrs.fr/eiss6/maekawa-eiss6.pdf, pp. 227–247.

McCarthy, John, and Alan Prince (1993), “Generalized Alignment.” Yearbook of Morphology : 79–153.McCloskey, James (2006), “Questions and Questioning in a Local English.” In Raffaella Zanuttini, Héctor Campos,

Elena Herburger, and Paul H. Portner, eds., Negation, Tense and Clausal Architecture: Cross-Linguistic Investiga-tions, Georgetown, D.C.: Georgetown University Press, pp. 86–126.

Pintzuk, Susan (1993), “Verb Seconding in Old English: Verb Movement to Infl.” The Linguistic Review 10: 5–35.Pintzuk, Susan (1999), Phrase Structures in Competition: Variation and Change in Old English Word Order. New York

and London: Garland.Platzack, Christer (1986), “COMP, INFL, and Germanic Word Order.” In Lars Hellan and Kirsti Koch Christensen, eds.,

Topics in Scandinavian Syntax, Dordrecht: Reidel, pp. 185–234.

21

Reinhart, Tanya (1981), “A Second Comp Position.” In Adriana Belletti, Luciana Brandi, and Luigi Rizzi, eds., Theory ofMarkedness in Generative Grammar: Proceedings of the 1979 GLOW Conference, Pisa: Scuola Normale Superioredi Pisa, pp. 517–557.

Rizzi, Luigi (1997), “The Fine Structure of the Left Periphery.” In Liliane Haegeman, ed., Elements of Grammar,Dordrecht: Kluwer, pp. 281–337.

Rögnvaldsson, Eirikur, and Höskuldur Thráinsson (1990), “On Icelandic Word Order Once More.” In Joan Maling andAnnie Zaenen, eds., Modern Icelandic Syntax, San Diego: Academic Press, pp. 3–41.

Santorini, Beatrice (1992), “Variation and Change in Yiddish Subordinate Clause Word Order.” Natural Language andLinguistic Theory 10: 595–640.

Sobin, Nicholas (2003), “Negative Inversion as Nonmovement.” Syntax 6: 183–212.Struckmeier, Volker (2014), “Ja doch wohl C? Modal Particles in German as C-Related Elements.” Studia Linguistica

68: 16–48.Vikner, Sten (1995), Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University

Press.Webelhuth, Gert (1992), Principles and Parameters of Syntactic Saturation. Oxford: Oxford University Press.

22